v0.1.0 — Stable Release

[0.1.0] - 2026-04-15

Fixed

• pyproject.toml: project.dependencies was incorrectly formatted as a TOML table instead of an array of PEP 508 strings, preventing installation (pip install failed with configuration error).
• Regression detection: SurrogateExtractor and Converter used a fragile _estimator_type instance-attribute check that failed for ensemble regressors (RandomForestRegressor, GradientBoostingRegressor) in scikit-learn ≥ 1.8. Replaced with sklearn.base.is_regressor() with a safe fallback for non-conformant mock models.
• Build system: Removed setuptools_scm from build dependencies (was unused; no git tag versioning configured) and dropped legacy setup.py that duplicated pyproject.toml metadata.

Changed

• Promoted from beta (0.1.0b1) to stable (0.1.0).
• Development Status classifier updated to 5 - Production/Stable.
• Added [tool.setuptools.packages.find] section to pyproject.toml for explicit package discovery.

Documentation (post-release)

• Added 6 Jupyter notebook examples (notebooks/) covering quickstart, classification, regression, feature analysis, multi-format export, and a full end-to-end IoT pipeline on the ADL Air Quality dataset.
• Integrated notebooks into MkDocs docs via mkdocs-jupyter plugin with a pre-build hook.
• Updated installation instructions across all docs: pip install blackbox2c as primary, with virtual environment guidance.
• Fixed docs/api/analysis.md plot() signature, optimize_rules description in config.md.
• Removed stale "PyPI release" item from roadmap (already released).

Infrastructure (post-release)

• Added publish.yml GitHub Actions workflow: tag-triggered, runs tests → build → PyPI Trusted Publisher → GitHub Release.
• Configured PyPI Trusted Publisher (OIDC) — no API token stored in secrets.